We present work on H alpha spectral line characteristics in PHOENIX stellar model atmospheres and their comparison to microlensing observations. We examine in detail the H alpha equivalent width (EW) and the line shape characteristics for effective temperatures of 4500K< Teff < 5600K where H alpha is a strong spectral feature. We find that H alpha EW in models calculated under the assumption of local thermodynamic equilibrium (LTE) is up to 15% smaller than in models without this assumption, non-LTE models (NLTE) and that line shapes vary significantly for the two model types. A comparison with available high quality microlensing data, capable of tracing H alpha absorption across the face of one G5III giant, shows that the LTE model that fits the EW best is about 100K hotter than and the best-fitting NLTE model has a similar Teff as predicted by the spectral type analysis of the observed star but agree within the uncertainties of the observationally derived temperature. Neither LTE nor NLTE models fit the line shape well. We suspect unmodelled chromospheric emission. Line shape diagnostics suggest lower gravities than derived for the star and are unacceptable low in the case of the LTE models. We show that EW alone is insufficient for comparison to stellar model atmospheres, but combined with a new shape parameter we define is promising. In stellar parameter ranges where the H alpha line is strong, a NLTE approach of modeling stellar atmospheres is not only beneficial but mandatory.

Exoplanetary transits produce a double-horned color signature that is distinct from both binaries and blends and can thus be used to separate exoplanets from false positives in transit searches. Color photometry with precision sufficient to detect this signal in transits of HD 209458 is available in the literature. Analysis of these observations reveals that, while the signature does exhibit the expected shape, it is significantly stronger than PHOENIX atmospheric models predict.

We analyze the photometric data obtained by PLANET and OGLE on the caustic-crossing binary-lens microlensing event OGLE-2002-BLG-069. Thanks to the excellent photometric and spectroscopic coverage of the event, we are able to constrain the lens model up to the known ambiguity between close and wide binary lenses. The detection of annual parallax in combination with measurements of extended-source effects allows us to determine the mass, distance and velocity of the lens components for the competing models. While the model involving a close binary lens leads to a Bulge-Disc lens scenario with a lens mass of M=(0.51 +- 0.15) M_sol and distance of D_L=(2.9 +- 0.4) kpc, the wide binary lens solution requires a rather implausible binary black-hole lens (M >=126 M_sol). Furthermore we compare current state-of-the-art numerical and empirical models for the surface brightness profile of the source, a G5III Bulge giant. We find that a linear limb-darkening model for the atmosphere of the source star is consistent with the data whereas a PHOENIX atmosphere model assuming LTE and with no free parameter does not match our observations.

We discuss high-resolution, time-resolved spectra of the caustic exit of the binary microlensing event OGLE 2002-BUL-69 obtained with UVES on the VLT. The source star is a G5III giant in the Galactic Bulge. During such events, the source star is highly magnified, and a strong differential magnification around the caustic resolves its surface. Using an appropriate model stellar atmosphere generated by the NextGEN code we obtained a model light curve for the caustic exit and compared it with a dense set of photometric observations obtained by the PLANET microlensing follow up network. We further compared predicted variations in the H alpha equivalent width with those measured from our spectra. While the model and observations agree in the gross features, there are discrepancies suggesting shortcomings in the model, particularly for the H alpha line core, where we have detected amplified emission from the stellar chromosphere as the source star's trailing limb exited the caustic. This achievement became possible by the provision of the OGLE-III Early Warning System, a network of small telescopes capable of nearly-continuous round-the-clock photometric monitoring, on-line data reduction, daily near-real-time modelling in order to predict caustic crossing parameters, and a fast and efficient response of a 8m-class telescope to a ``Target-Of-Opportunity'' observation request.